Digital Signal Processing Reference
In-Depth Information
If the noise power in a subband is increased, the magnitude-error on a demod-
ulated qam constellation point will also increase. In the constellation diagram,
an error 'cloud' appears around the ideal constellation point (Figure 1.8). If
the deviation from the original point becomes larger than half the distance
between two constellation points, the (hard-decision) qam-demapper of the
receiver makes a wrong decision and decodes the incorrect symbol. Now con-
sider the case where the snr of the wireless channel is just enough to allow
an 802
.
11a/g link to operate at its maximum data rate, being 54 Mbps using a
64-qam modulation scheme [Wla07].
Suddenly, a narrowband interferer descends from the blue sky. Suppose that
the power of this divine interferer is comparable to the transmission power of
the ofdm link, but the distance between the source of interference and the re-
ceiver is half of the distance of the ofdm link. It was shown earlier that the
receiver had to give up 1 bit of the accuracy of the ad-converter, only to be able
to cope with the power of the spur. As a result of the loss of 1 bit, simulations
show that the implementation loss (il) due to the reduced signal-to-noise ratio
is 0
.
37 dB ([Eng02] p. 129). Since the link quality was assumed barely enough
to sustain the current data rate, the 802
.
11a/g link controller has three options:
decrease the coding rate (R), reduce the modulation depth, or just fail.
The 802
.
11a/g standard uses two convolutional coding schemes at the 64-qam
modulation depth. For the maximum throughput of 54 Mbps, a coding rate of
R
=
3
/
4 provides a coding gain of 5
.
7 dB. If the coding rate is decreased
to R
2
/
3, the data rate drops to 48 Mbps in exchange for a coding gain
of 6
.
0 dB. It should be clear that the additional coding gain of the latter case
(0
.
3 dB) is not sufficient to compensate for the implementation loss of 0
.
37 dB.
So the link controller is forced to take more aggressive measures to counter
the reduced link quality: reduce the modulation depth. From (1.1), it follows
that a fallback from 64-qam to 16qam increases the msed by 6 dB. This
means that up to four times more noise energy can be tolerated compared to
the 64-qam reference case, more than sufficient to compensate for the 0
.
37 dB
implementation loss in the ad-converter of the receiver. The downside of all
this is that the throughput of the 802
.
11a/g link is being reduced by more than
30%, from 54 to 36 Mbps, even for the fairly conservative case where the in-
band interferer power is only four times larger than the average ofdm power.
The previous example shows that in-band interference has a serious impact
on the sensitivity of a wideband receiver, if it cannot be properly removed
from the signal chain in an early stage. In order to get some intuitive feeling for
orders of magnitude, the interference immunity of ofdm-based wireless lan
is put in perspective to the blocking specifications of the gsm-900 standard
(Figure 1.9).
=
